2 * Copyright (c) 2003, 2004 Jeffrey M. Hsu. All rights reserved.
3 * Copyright (c) 2003, 2004 The DragonFly Project. All rights reserved.
5 * This code is derived from software contributed to The DragonFly Project
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of The DragonFly Project nor the names of its
17 * contributors may be used to endorse or promote products derived
18 * from this software without specific, prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
36 * The Regents of the University of California. All rights reserved.
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by the University of
49 * California, Berkeley and its contributors.
50 * 4. Neither the name of the University nor the names of its contributors
51 * may be used to endorse or promote products derived from this software
52 * without specific prior written permission.
54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
66 * @(#)tcp_timer.c 8.2 (Berkeley) 5/24/95
67 * $FreeBSD: src/sys/netinet/tcp_timer.c,v 1.34.2.14 2003/02/03 02:33:41 hsu Exp $
68 * $DragonFly: src/sys/netinet/tcp_timer.c,v 1.17 2008/03/30 20:39:01 dillon Exp $
71 #include "opt_compat.h"
72 #include "opt_inet6.h"
73 #include "opt_tcpdebug.h"
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/kernel.h>
79 #include <sys/sysctl.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/protosw.h>
83 #include <sys/thread.h>
84 #include <sys/globaldata.h>
85 #include <sys/thread2.h>
86 #include <sys/msgport2.h>
88 #include <machine/cpu.h> /* before tcp_seq.h, for tcp_random18() */
90 #include <net/route.h>
91 #include <net/netmsg2.h>
93 #include <netinet/in.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_pcb.h>
99 #include <netinet/ip_var.h>
100 #include <netinet/tcp.h>
101 #include <netinet/tcp_fsm.h>
102 #include <netinet/tcp_seq.h>
103 #include <netinet/tcp_timer.h>
104 #include <netinet/tcp_timer2.h>
105 #include <netinet/tcp_var.h>
106 #include <netinet/tcpip.h>
108 #include <netinet/tcp_debug.h>
111 #define TCP_TIMER_REXMT 0x01
112 #define TCP_TIMER_PERSIST 0x02
113 #define TCP_TIMER_KEEP 0x04
114 #define TCP_TIMER_2MSL 0x08
115 #define TCP_TIMER_DELACK 0x10
117 static struct tcpcb *tcp_timer_rexmt_handler(struct tcpcb *);
118 static struct tcpcb *tcp_timer_persist_handler(struct tcpcb *);
119 static struct tcpcb *tcp_timer_keep_handler(struct tcpcb *);
120 static struct tcpcb *tcp_timer_2msl_handler(struct tcpcb *);
121 static struct tcpcb *tcp_timer_delack_handler(struct tcpcb *);
123 static const struct tcp_timer {
125 struct tcpcb *(*tt_handler)(struct tcpcb *);
126 } tcp_timer_handlers[] = {
127 { TCP_TIMER_DELACK, tcp_timer_delack_handler },
128 { TCP_TIMER_REXMT, tcp_timer_rexmt_handler },
129 { TCP_TIMER_PERSIST, tcp_timer_persist_handler },
130 { TCP_TIMER_KEEP, tcp_timer_keep_handler },
131 { TCP_TIMER_2MSL, tcp_timer_2msl_handler },
136 sysctl_msec_to_ticks(SYSCTL_HANDLER_ARGS)
140 tt = *(int *)oidp->oid_arg1;
141 s = (int)((int64_t)tt * 1000 / hz);
143 error = sysctl_handle_int(oidp, &s, 0, req);
144 if (error || !req->newptr)
147 tt = (int)((int64_t)s * hz / 1000);
151 *(int *)oidp->oid_arg1 = tt;
156 SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINIT, keepinit, CTLTYPE_INT|CTLFLAG_RW,
157 &tcp_keepinit, 0, sysctl_msec_to_ticks, "I", "Time to establish TCP connection");
160 SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPIDLE, keepidle, CTLTYPE_INT|CTLFLAG_RW,
161 &tcp_keepidle, 0, sysctl_msec_to_ticks, "I", "Time before TCP keepalive probes begin");
164 SYSCTL_PROC(_net_inet_tcp, TCPCTL_KEEPINTVL, keepintvl, CTLTYPE_INT|CTLFLAG_RW,
165 &tcp_keepintvl, 0, sysctl_msec_to_ticks, "I", "Time between TCP keepalive probes");
168 SYSCTL_PROC(_net_inet_tcp, TCPCTL_DELACKTIME, delacktime,
169 CTLTYPE_INT|CTLFLAG_RW, &tcp_delacktime, 0, sysctl_msec_to_ticks, "I",
170 "Time before a delayed ACK is sent");
173 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, msl, CTLTYPE_INT|CTLFLAG_RW,
174 &tcp_msl, 0, sysctl_msec_to_ticks, "I", "Maximum segment lifetime");
177 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_min, CTLTYPE_INT|CTLFLAG_RW,
178 &tcp_rexmit_min, 0, sysctl_msec_to_ticks, "I", "Minimum Retransmission Timeout");
181 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, rexmit_slop, CTLTYPE_INT|CTLFLAG_RW,
182 &tcp_rexmit_slop, 0, sysctl_msec_to_ticks, "I",
183 "Retransmission Timer Slop");
185 static int always_keepalive = 1;
186 SYSCTL_INT(_net_inet_tcp, OID_AUTO, always_keepalive, CTLFLAG_RW,
187 &always_keepalive , 0, "Assume SO_KEEPALIVE on all TCP connections");
189 /* max idle probes */
190 int tcp_keepcnt = TCPTV_KEEPCNT;
191 SYSCTL_INT(_net_inet_tcp, OID_AUTO, keepcnt, CTLFLAG_RW,
192 &tcp_keepcnt, 0, "Maximum number of keepalive probes to be sent");
194 static int tcp_do_eifel_response = 1;
195 SYSCTL_INT(_net_inet_tcp, OID_AUTO, eifel_response, CTLFLAG_RW,
196 &tcp_do_eifel_response, 0, "Eifel response algorithm (RFC 4015)");
198 int tcp_eifel_rtoinc = 2;
199 SYSCTL_PROC(_net_inet_tcp, OID_AUTO, eifel_rtoinc, CTLTYPE_INT|CTLFLAG_RW,
200 &tcp_eifel_rtoinc, 0, sysctl_msec_to_ticks, "I",
201 "Eifel response RTO increment");
203 /* max idle time in persist */
204 int tcp_maxpersistidle;
207 * Cancel all timers for TCP tp.
210 tcp_canceltimers(struct tcpcb *tp)
212 tcp_callout_stop(tp, tp->tt_2msl);
213 tcp_callout_stop(tp, tp->tt_persist);
214 tcp_callout_stop(tp, tp->tt_keep);
215 tcp_callout_stop(tp, tp->tt_rexmt);
219 * Caller should be in critical section
222 tcp_send_timermsg(struct tcpcb *tp, uint32_t task)
224 struct netmsg_tcp_timer *tmsg = tp->tt_msg;
226 KKASSERT(tmsg != NULL && tmsg->tt_cpuid == mycpuid &&
227 tmsg->tt_tcb != NULL);
229 tmsg->tt_tasks |= task;
230 if (tmsg->tt_msg.lmsg.ms_flags & MSGF_DONE)
231 lwkt_sendmsg(tmsg->tt_msgport, &tmsg->tt_msg.lmsg);
234 int tcp_syn_backoff[TCP_MAXRXTSHIFT + 1] =
235 { 1, 1, 1, 1, 1, 2, 4, 8, 16, 32, 64, 64, 64 };
237 int tcp_syn_backoff_low[TCP_MAXRXTSHIFT + 1] =
238 { 1, 1, 2, 4, 8, 8, 16, 16, 32, 64, 64, 64, 64 };
240 int tcp_backoff[TCP_MAXRXTSHIFT + 1] =
241 { 1, 2, 4, 8, 16, 32, 64, 64, 64, 64, 64, 64, 64 };
243 static int tcp_totbackoff = 511; /* sum of tcp_backoff[] */
245 /* Caller should be in critical section */
246 static struct tcpcb *
247 tcp_timer_delack_handler(struct tcpcb *tp)
249 tp->t_flags |= TF_ACKNOW;
250 tcpstat.tcps_delack++;
256 * TCP timer processing.
259 tcp_timer_delack(void *xtp)
261 struct tcpcb *tp = xtp;
262 struct callout *co = &tp->tt_delack->tc_callout;
265 if (callout_pending(co) || !callout_active(co)) {
269 callout_deactivate(co);
270 tcp_send_timermsg(tp, TCP_TIMER_DELACK);
274 /* Caller should be in critical section */
275 static struct tcpcb *
276 tcp_timer_2msl_handler(struct tcpcb *tp)
283 ostate = tp->t_state;
286 * 2 MSL timeout in shutdown went off. If we're closed but
287 * still waiting for peer to close and connection has been idle
288 * too long, or if 2MSL time is up from TIME_WAIT, delete connection
289 * control block. Otherwise, check again in a bit.
291 if (tp->t_state != TCPS_TIME_WAIT &&
292 (ticks - tp->t_rcvtime) <= tp->t_maxidle) {
293 tcp_callout_reset(tp, tp->tt_2msl, tp->t_keepintvl,
300 if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
301 tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
307 tcp_timer_2msl(void *xtp)
309 struct tcpcb *tp = xtp;
310 struct callout *co = &tp->tt_2msl->tc_callout;
313 if (callout_pending(co) || !callout_active(co)) {
317 callout_deactivate(co);
318 tcp_send_timermsg(tp, TCP_TIMER_2MSL);
322 /* Caller should be in critical section */
323 static struct tcpcb *
324 tcp_timer_keep_handler(struct tcpcb *tp)
326 struct tcptemp *t_template;
328 int ostate = tp->t_state;
332 * Keep-alive timer went off; send something
333 * or drop connection if idle for too long.
335 tcpstat.tcps_keeptimeo++;
336 if (tp->t_state < TCPS_ESTABLISHED)
338 if ((always_keepalive || (tp->t_flags & TF_KEEPALIVE) ||
339 (tp->t_inpcb->inp_socket->so_options & SO_KEEPALIVE)) &&
340 tp->t_state <= TCPS_CLOSING) {
341 if ((ticks - tp->t_rcvtime) >= tp->t_keepidle + tp->t_maxidle)
344 * Send a packet designed to force a response
345 * if the peer is up and reachable:
346 * either an ACK if the connection is still alive,
347 * or an RST if the peer has closed the connection
348 * due to timeout or reboot.
349 * Using sequence number tp->snd_una-1
350 * causes the transmitted zero-length segment
351 * to lie outside the receive window;
352 * by the protocol spec, this requires the
353 * correspondent TCP to respond.
355 tcpstat.tcps_keepprobe++;
356 t_template = tcp_maketemplate(tp);
358 tcp_respond(tp, t_template->tt_ipgen,
359 &t_template->tt_t, NULL,
360 tp->rcv_nxt, tp->snd_una - 1, 0);
361 tcp_freetemplate(t_template);
363 tcp_callout_reset(tp, tp->tt_keep, tp->t_keepintvl,
366 tcp_callout_reset(tp, tp->tt_keep, tp->t_keepidle,
371 if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
372 tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
377 tcpstat.tcps_keepdrops++;
378 tp = tcp_drop(tp, ETIMEDOUT);
381 if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
382 tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
388 tcp_timer_keep(void *xtp)
390 struct tcpcb *tp = xtp;
391 struct callout *co = &tp->tt_keep->tc_callout;
394 if (callout_pending(co) || !callout_active(co)) {
398 callout_deactivate(co);
399 tcp_send_timermsg(tp, TCP_TIMER_KEEP);
403 /* Caller should be in critical section */
404 static struct tcpcb *
405 tcp_timer_persist_handler(struct tcpcb *tp)
412 ostate = tp->t_state;
415 * Persistance timer into zero window.
416 * Force a byte to be output, if possible.
418 tcpstat.tcps_persisttimeo++;
420 * Hack: if the peer is dead/unreachable, we do not
421 * time out if the window is closed. After a full
422 * backoff, drop the connection if the idle time
423 * (no responses to probes) reaches the maximum
424 * backoff that we would use if retransmitting.
426 if (tp->t_rxtshift == TCP_MAXRXTSHIFT &&
427 ((ticks - tp->t_rcvtime) >= tcp_maxpersistidle ||
428 (ticks - tp->t_rcvtime) >= TCP_REXMTVAL(tp) * tcp_totbackoff)) {
429 tcpstat.tcps_persistdrop++;
430 tp = tcp_drop(tp, ETIMEDOUT);
434 tp->t_flags |= TF_FORCE;
436 tp->t_flags &= ~TF_FORCE;
440 if (tp && tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
441 tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
447 tcp_timer_persist(void *xtp)
449 struct tcpcb *tp = xtp;
450 struct callout *co = &tp->tt_persist->tc_callout;
453 if (callout_pending(co) || !callout_active(co)){
457 callout_deactivate(co);
458 tcp_send_timermsg(tp, TCP_TIMER_PERSIST);
463 tcp_save_congestion_state(struct tcpcb *tp)
465 tp->snd_cwnd_prev = tp->snd_cwnd;
466 tp->snd_wacked_prev = tp->snd_wacked;
467 tp->snd_ssthresh_prev = tp->snd_ssthresh;
468 tp->snd_recover_prev = tp->snd_recover;
470 tp->t_rxtcur_prev = tp->t_rxtcur;
471 tp->t_srtt_prev = tp->t_srtt +
472 (tcp_eifel_rtoinc << TCP_RTT_SHIFT);
473 tp->t_rttvar_prev = tp->t_rttvar;
474 tp->snd_max_prev = tp->snd_max;
475 tp->rxt_flags &= ~TRXT_F_REBASERTO;
477 if (IN_FASTRECOVERY(tp))
478 tp->rxt_flags |= TRXT_F_WASFRECOVERY;
480 tp->rxt_flags &= ~TRXT_F_WASFRECOVERY;
481 if (tp->t_flags & TF_RCVD_TSTMP) {
482 tp->t_rexmtTS = ticks;
483 tp->rxt_flags |= TRXT_F_FIRSTACCACK;
486 tcp_sack_save_scoreboard(&tp->scb);
491 tcp_revert_congestion_state(struct tcpcb *tp)
493 tp->snd_cwnd = tp->snd_cwnd_prev;
494 tp->snd_wacked = tp->snd_wacked_prev;
495 tp->snd_ssthresh = tp->snd_ssthresh_prev;
496 tp->snd_recover = tp->snd_recover_prev;
497 if (tp->rxt_flags & TRXT_F_WASFRECOVERY)
498 ENTER_FASTRECOVERY(tp);
499 if (tp->rxt_flags & TRXT_F_FASTREXMT) {
500 ++tcpstat.tcps_sndfastrexmitbad;
501 if (tp->rxt_flags & TRXT_F_EARLYREXMT)
502 ++tcpstat.tcps_sndearlyrexmitbad;
504 ++tcpstat.tcps_sndrtobad;
505 tp->snd_last = ticks;
506 if (tcp_do_eifel_response)
507 tp->rxt_flags |= TRXT_F_REBASERTO;
511 tp->snd_nxt = tp->snd_max;
513 tcp_sack_revert_scoreboard(&tp->scb, tp->snd_una);
517 /* Caller should be in critical section */
518 static struct tcpcb *
519 tcp_timer_rexmt_handler(struct tcpcb *tp)
527 ostate = tp->t_state;
530 * Retransmission timer went off. Message has not
531 * been acked within retransmit interval. Back off
532 * to a longer retransmit interval and retransmit one segment.
534 if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
535 tp->t_rxtshift = TCP_MAXRXTSHIFT;
536 tcpstat.tcps_timeoutdrop++;
537 tp = tcp_drop(tp, tp->t_softerror ?
538 tp->t_softerror : ETIMEDOUT);
541 if (tp->t_rxtshift == 1) {
543 * first retransmit; record ssthresh and cwnd so they can
544 * be recovered if this turns out to be a "bad" retransmit.
545 * A retransmit is considered "bad" if an ACK for this
546 * segment is received within RTT/2 interval; the assumption
547 * here is that the ACK was already in flight. See
548 * "On Estimating End-to-End Network Path Properties" by
549 * Allman and Paxson for more details.
551 tp->t_badrxtwin = ticks + (tp->t_srtt >> (TCP_RTT_SHIFT + 1));
552 tcp_save_congestion_state(tp);
553 tp->rxt_flags &= ~(TRXT_F_FASTREXMT | TRXT_F_EARLYREXMT);
555 if (tp->t_state == TCPS_SYN_SENT || tp->t_state == TCPS_SYN_RECEIVED) {
557 * Record the time that we spent in SYN or SYN|ACK
560 * Needed by RFC3390 and RFC6298.
562 tp->t_rxtsyn += tp->t_rxtcur;
564 /* Throw away SACK blocks on a RTO, as specified by RFC2018. */
565 tcp_sack_cleanup(&tp->scb);
566 tcpstat.tcps_rexmttimeo++;
567 if (tp->t_state == TCPS_SYN_SENT) {
568 if (tcp_low_rtobase) {
569 rexmt = TCP_REXMTVAL(tp) *
570 tcp_syn_backoff_low[tp->t_rxtshift];
572 rexmt = TCP_REXMTVAL(tp) *
573 tcp_syn_backoff[tp->t_rxtshift];
576 rexmt = TCP_REXMTVAL(tp) * tcp_backoff[tp->t_rxtshift];
578 TCPT_RANGESET(tp->t_rxtcur, rexmt,
579 tp->t_rttmin, TCPTV_REXMTMAX);
581 * If losing, let the lower level know and try for
582 * a better route. Also, if we backed off this far,
583 * our srtt estimate is probably bogus. Clobber it
584 * so we'll take the next rtt measurement as our srtt;
585 * move the current srtt into rttvar to keep the current
586 * retransmit times until then.
588 if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
590 if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0)
591 in6_losing(tp->t_inpcb);
594 in_losing(tp->t_inpcb);
595 tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
598 tp->snd_nxt = tp->snd_una;
599 tp->rexmt_high = tp->snd_una;
600 tp->sack_flags &= ~TSACK_F_SACKRESCUED;
601 tp->snd_recover = tp->snd_max;
603 * Force a segment to be sent.
605 tp->t_flags |= TF_ACKNOW;
607 * If timing a segment in this window, stop the timer.
611 * Close the congestion window down to one segment
612 * (we'll open it by one segment for each ack we get).
613 * Since we probably have a window's worth of unacked
614 * data accumulated, this "slow start" keeps us from
615 * dumping all that data as back-to-back packets (which
616 * might overwhelm an intermediate gateway).
618 * There are two phases to the opening: Initially we
619 * open by one mss on each ack. This makes the window
620 * size increase exponentially with time. If the
621 * window is larger than the path can handle, this
622 * exponential growth results in dropped packet(s)
623 * almost immediately. To get more time between
624 * drops but still "push" the network to take advantage
625 * of improving conditions, we switch from exponential
626 * to linear window opening at some threshhold size.
627 * For a threshhold, we use half the current window
628 * size, truncated to a multiple of the mss.
630 * (the minimum cwnd that will give us exponential
631 * growth is 2 mss. We don't allow the threshhold
635 u_int win = min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
639 tp->snd_cwnd = tp->t_maxseg;
641 tp->snd_ssthresh = win * tp->t_maxseg;
644 EXIT_FASTRECOVERY(tp);
649 if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
650 tcp_trace(TA_USER, ostate, tp, NULL, NULL, PRU_SLOWTIMO);
656 tcp_timer_rexmt(void *xtp)
658 struct tcpcb *tp = xtp;
659 struct callout *co = &tp->tt_rexmt->tc_callout;
662 if (callout_pending(co) || !callout_active(co)) {
666 callout_deactivate(co);
667 tcp_send_timermsg(tp, TCP_TIMER_REXMT);
672 tcp_timer_handler(netmsg_t msg)
674 struct netmsg_tcp_timer *tmsg = (struct netmsg_tcp_timer *)msg;
675 const struct tcp_timer *tt;
680 KKASSERT(tmsg->tt_cpuid == mycpuid && tmsg->tt_tcb != NULL);
683 /* Save pending tasks and reset the tasks in message */
684 tmsg->tt_running_tasks = tmsg->tt_tasks;
685 tmsg->tt_prev_tasks = tmsg->tt_tasks;
689 lwkt_replymsg(&tmsg->tt_msg.lmsg, 0);
691 if (tmsg->tt_running_tasks == 0) {
693 * All of the timers are cancelled when the message
694 * is pending; bail out.
700 for (tt = tcp_timer_handlers; tt->tt_handler != NULL; ++tt) {
701 if ((tmsg->tt_running_tasks & tt->tt_task) == 0)
704 tmsg->tt_running_tasks &= ~tt->tt_task;
705 tp = tt->tt_handler(tp);
709 if (tmsg->tt_running_tasks == 0) /* nothing left to do */
717 tcp_create_timermsg(struct tcpcb *tp, struct lwkt_port *msgport)
719 struct netmsg_tcp_timer *tmsg = tp->tt_msg;
721 netmsg_init(&tmsg->tt_msg, NULL, &netisr_adone_rport,
722 MSGF_DROPABLE | MSGF_PRIORITY, tcp_timer_handler);
723 tmsg->tt_cpuid = mycpuid;
724 tmsg->tt_msgport = msgport;
730 tcp_destroy_timermsg(struct tcpcb *tp)
732 struct netmsg_tcp_timer *tmsg = tp->tt_msg;
734 if (tmsg == NULL || /* listen socket */
735 tmsg->tt_tcb == NULL) /* only tcp_attach() is called */
738 KKASSERT(tmsg->tt_cpuid == mycpuid);
740 if ((tmsg->tt_msg.lmsg.ms_flags & MSGF_DONE) == 0) {
742 * This message is still pending to be processed;
745 lwkt_dropmsg(&tmsg->tt_msg.lmsg);
751 tcp_callout_init(struct tcp_callout *tc, uint32_t task)
753 callout_init_mp(&tc->tc_callout);
758 tcp_inittimers(struct tcpcb *tp)
760 tcp_callout_init(tp->tt_rexmt, TCP_TIMER_REXMT);
761 tcp_callout_init(tp->tt_persist, TCP_TIMER_PERSIST);
762 tcp_callout_init(tp->tt_keep, TCP_TIMER_KEEP);
763 tcp_callout_init(tp->tt_2msl, TCP_TIMER_2MSL);
764 tcp_callout_init(tp->tt_delack, TCP_TIMER_DELACK);